Model-based Fault Detection for a Turbocharger

Turbochargers are widely used on automotive, power generation and marine applications. However, while they are
being used, several types of faults such as rotor unbalance, journal bearings abrasion, insufficient oil supply, failure
from excessive exhaust temperatures, stable and stall, etc might happen. It would be a time and money consuming
work, even sometimes impossible to inspect periodically. In order to detect and classify those kinds of faults
effectively, this paper tries to apply a model-based approach on turbocharger monitoring. Firstly, a mathematic model
for a turbocharger is developed in Rotor-dynamics considering the effects of gyroscopic and journal bearing. Secondly,
the model of the whole system is validated by relevant experiments. Thirdly, data processing and faults classification
techniques are used to analyze the residual signals between those collected from faulting turbochargers and simulated
by the model. Finally, a conclusion will be drawn on the performance of this approach on the specific application.